I teach engineering, AP Computer Science, and AP Physics, and for years the assignments I put off longest were never the essays. They were the recordings. A two-minute video of a student explaining how their circuit works. An audio file of a kid reading their capstone reflection out loud. A screen recording of someone walking through their code. That is some of the best evidence I get of what a student actually understood, and it sat in a folder I could not get through fast enough.
Written work has a shortcut. You can skim an essay and know within a paragraph roughly where it lands. You cannot skim a video. You have to watch it in real time, and thirty submissions at three minutes each is an hour and a half before you have written a single comment. That arithmetic is why a lot of us assign less spoken and recorded work than we know we should. Here is how I grade it now without giving up the final say on anyone's grade.
Why spoken and recorded work is hard to grade at scale
The problem is not that recordings are hard to judge. Most of us can tell a strong oral presentation from a weak one in the first thirty seconds. The problem is throughput. An essay lets you read at your own pace and jump around; a recording forces you into the student's pace, front to back, with no scanning.
Spoken work also resists a paper trail. You finish watching, the evidence is gone from the screen, and the note you write is from memory. By the twentieth submission your memory is tired and your scores drift. So most of us make a quiet tradeoff: assign fewer recordings, or grade them on a loose gut-check instead of a rubric. Both are a loss. Spoken and performance tasks are often the truest read on whether a student learned the thing, especially in world language, science, and any subject where explaining your work is the point.
What ClassLens can grade
Most AI grading tools only handle text. You paste an essay, you get a score. The moment the assignment is a spoken response or a recorded demonstration, they go quiet.
ClassLens grades every assignment type, and since June 2026 that includes audio and video submissions alongside essays, short answers, DBQs, worksheets, quizzes, and programming work. If a student attached it to a Google Classroom assignment (an audio file, a video, a Drive recording), ClassLens can pull it and grade it against the rubric you set. The common formats are covered, MP3, M4A, WAV, MP4, MOV, and WEBM among them. Anything it cannot read, like the MKV that OBS records by default, comes back marked as skipped rather than guessed at, so you know to ask for a re-export instead of finding out from a wrong grade.
Here is the recorded work teachers actually bring to it:
- World-language speaking practice and oral proficiency checks
- Oral presentations and project defenses
- Science explanations and lab talk-throughs
- Reading-fluency recordings
- Speech and debate rounds
- Music performance
- PE skill demonstrations
- Video book reports
The common thread is a rubric. If you can describe what a strong response sounds or looks like, ClassLens can draft a grade and feedback against it for every student, then hand the pile back for you to check.
How to grade a video or audio assignment in Google Classroom
The flow is the same one you would use for an essay. Nothing gets uploaded or pasted by hand. ClassLens reads the submissions straight from Classroom.
1. Sign in with Google
Use the same Google account you use for Classroom. You can start from classlens.com or install from the Google Workspace Marketplace.
2. Pick the class and the assignment
ClassLens lists your Google Classroom classes and the assignments inside them. Choose the speaking task, the presentation, the lab video, whatever your students recorded and turned in.
3. Set your rubric and preferences
Paste in your rubric, or write the criteria directly, and set your grading preferences: strictness, how much feedback you want, your late-work policy. This is the step that matters most for spoken work, and there are tips for it below.
4. Run the batch
ClassLens pulls each student's recording and grades it against your rubric through Google Cloud Vertex AI. It drafts a score and a short justification per submission, and it builds a class-level Knowledge Gap Report so you can see the misconceptions showing up across the whole class, not just one kid at a time.
5. Review every grade, then return them
Nothing reaches a student automatically. Every draft lands in the Batch Review Dashboard, where you read the AI's grade and comment, change anything you disagree with, and then either save the grades as drafts in Classroom or click to return the reviewed batch to students in one pass.
Want to see what the output looks like before you connect an account? The public demo shows a graded sample class and its Knowledge Gap Report with no login. The sample assignment there is written work, not a recording, but the review screen is the same one your recordings land in.
Writing a rubric for spoken and recorded work
A rubric written for an essay will underperform on a recording, because it usually scores things you cannot hear. A few adjustments help.
Score what is in the recording, not the production. Unless video editing is the skill you are teaching, keep camera quality, background, and polish out of the rubric. Grade the content of what the student said or demonstrated. If you leave "professional presentation" in as a criterion, be specific about whether you mean the argument or the lighting.
Name the spoken behaviors you want. For a world-language speaking check, that might be pronunciation, sentence complexity, and vocabulary range. For a science explanation, it might be correct use of terms and whether the student names their evidence. The more concrete the criterion, the closer the AI draft lands to your own read.
Separate delivery from understanding. A quiet student who explains a concept perfectly and a confident student who explains it wrong are two different grades. If both matter to you, make them two lines on the rubric so the score does not blur them together.
Set the anchors. Hand-grade three recordings yourself first, one weak, one middle, one strong, so you calibrate your own sense of the rubric before you run the batch. Then the AI draft is something you check against a standard you set, not something you trust blind.
Staying in control: you review every grade
ClassLens does not return grades on its own. Auto-return was removed from the product in April 2026, and there is no mode where an AI-generated grade reaches a student without a teacher looking at it first.
That is deliberate. The Batch Review Dashboard lets the AI do the mechanical pass, listening to every recording and drafting a score and comment against your rubric, while you do the judgment pass. You read down the batch, fix what is wrong, and release what is right. On a class of thirty, that turns ninety minutes of watching into a review you can finish over a coffee, without giving up the final call on anyone's grade.
Where the AI helps, and where it doesn't
The honest version. ClassLens is good at the high-volume, rubric-shaped part of grading recordings. It listens to every submission at full attention, applies the same standard to the last student as the first, and never gets tired at submission twenty. For reading fluency, speaking checks, lab explanations, and presentations scored against a clear rubric, that is most of the work.
It is weaker on the parts that are not in the rubric. It will not read the room the way you do: the nervous kid who finally spoke up, or the growth since last month's recording that no rubric line captures. It grades against the criteria you give it, not against the year you have spent with this class. For anything high-stakes going onto a transcript, treat the AI draft as a second opinion, not the final word.
Disclosure: I am the founder of ClassLens and a working high school teacher. Take the product sections as a description of what the tool does, and test it against your own already-graded recordings before you trust it on a live batch. Where you disagree with the draft is exactly where your judgment is worth the most.
FAQ
Can AI grade a video assignment in Google Classroom?
Yes. ClassLens grades video and audio submissions turned in through Google Classroom, against a rubric you set. It drafts a score and feedback for each student and puts them in a review dashboard for you to check and return. Support for audio and video shipped in June 2026.
Do students have to upload their recordings somewhere else?
No. Students turn work in through Google Classroom the way they already do. ClassLens pulls the submissions directly from the assignment. There is no separate upload, and nothing to paste in by hand.
Does the AI return grades to students automatically?
No. Every grade is a draft until you review it. You read each one in the Batch Review Dashboard, edit what you want, and then either save the grades as drafts or return the reviewed batch yourself. Auto-return was removed from the product in April 2026.
What kinds of recorded work does it handle?
Speaking checks, oral presentations, lab explanations, reading fluency, music and PE performance, and video book reports are common. If you can write a rubric for it, ClassLens can draft grades against it.
What does it cost to try?
Every new account gets a 30-day full-access free trial, and the Free plan after that covers 100 submissions a month. Paid plans are Pro at $10 a month and Max at $20 a month. Full details are on the pricing page, and you can explore a graded sample with no login at the demo.
Try it on one assignment
Pick one recorded assignment you have been putting off: a stack of speaking checks, a set of lab videos, a folder of presentations. Set your rubric, run the batch, and read down the drafts in the review dashboard before you return a single grade. You will know inside one class period whether it saves you the time it promises.
Start free at classlens.com, or look at a graded sample first at the demo.